1. Field of the Invention
The present invention generally relates to a graphic contour extracting method, a pattern inspecting method, a program and a pattern inspecting system.
2. Description of the Prior Art
In a process of inspection in a semiconductor manufacturing process, image data of a pattern to be inspected are often acquired by an electron microscope, an optical microscope, a charge coupled device (CCD) or the like, to utilize contour information on the pattern to be inspected which is extracted on the basis of the image data. As contour in formation extracting methods, various methods have been put to practical use. For example, such methods include a method for setting a predetermined density value as a threshold value to define a portion corresponding to the threshold value as an edge when an acquired pattern image is a variable density image, a method for comparing an acquired pattern image with a closed curve graphic form serving as a reference, and a method for detecting a pattern edge by means of a two-dimensional filter, such as a Sobel filter.
Generally, when such a process is carried out, a region of interest (ROI) is assigned to a region of interest in an obtained image, and an edge extracting algorithm for executing any one of the above described various methods is executed in the interior of the assigned region to acquire a sequence of coordinate points of an edge corresponding to contour information to calculate geometric characteristic amounts, such as a line width, an area and a position of center of gravity, from data with respect to the obtained sequence of coordinate points.
As a shape of a boundary for defining an ROI region (an ROI boundary), a rectangle, an ellipse or a circle is often adopted for convenience of definition of the region. A method for assigning a rectangular ROI region is most generally carried out. An example of a method for adopting an elliptical ROI region is disclosed in e.g. Japanese Patent Laid-Open No. 2001-091231.
When the rectangular ROI boundary is adopted in the prior art, a pattern edge is searched in a direction parallel to any one of sides of an ROI region 62 as shown in, e.g., FIG. 27. Such a method is used when a pattern edge is substantially straight like a pattern P50 shown in the figure.
With respect to a pattern having a closed curved contour, an elliptical (or a circle) ROI boundary is set so as to surround a pattern edge of the pattern (P52) as shown by, e.g., an ROI boundary 64b in FIG. 28, and a pattern edge is searched in a direction of a radius vector.
However, in an image of a semiconductor pattern, the two-dimensional shape of a body serving as an object is more complicated. Therefore, in an inspecting method in the prior art, there are some cases where an edge searching direction is parallel to a pattern edge. For example, in the prior art, in the case of a pattern P2 shown in FIG. 29A, a pattern edge is locally parallel to an edge searching direction in a region Ep2 which includes an end of the pattern p2. In such a case, an edge point to be extracted has multi-values, so that there is a problem in that it is not possible to extract a proper edge position. Even with respect to a pattern having a closed curved contour, there may be regions Ep4, Ep6, Ep8 and Ep10 wherein pattern edges are parallel to edge searching directions (directions of radius vectors) as shown by a pattern P4 in FIG. 30A. In these cases, a tester must set a plurality of ROIs in accordance with the complexity of the pattern as shown by an ROI 68 in FIG. 29B and ROIs 74a through 74g in FIG. 30B, respectively, so that the costs required for inspection are increased. In particular, there are some cases where patterns to be inspected in fact have more complicated contour shapes as shown by patterns P6 and P8 in FIGS. 31 and 32, respectively. For example, with respect to the pattern P6 shown in FIG. 31, the edge searching direction locally corresponds to approximate directions of pattern edges at two portions (regions Ep12 and Ep14), so that there is a problem in that the tester fails in searching for edge positions in such portions. In the pattern P8 shown in FIG. 32, an elliptical ROI boundary 78b is set for searching edges in directions of its radius vectors. Therefore, the edge searching directions substantially corresponds to directions of pattern edges in regions Ep16 through Ep30, so that there is a problem in that the tester fails in searching for edge positions.
In addition, even if a pattern has a substantially straight pattern edge, when end portions of two patterns P10 and P12 face each other as shown in, e.g., FIG. 33, or when pattern edges of two close line patterns are perpendicular to each other as shown by patterns P14 and P18 in FIG. 34, there is a problem in that an edge searching direction substantially corresponds to directions of edges of a line pattern (P14) at two portions if search is carried out in a lateral direction on the figure.
Moreover, as shown by an example of a scanning electron microscope (SEM) image in FIG. 35, there are some cases where an OPC pattern called a serif is given to a part of a pattern on a reticle. In the case of a pattern P64 shown in this figure, it is not possible to detect edges in a contour portion parallel to an edge searching direction in, e.g., a region Ep32. There is also a problem in that a plurality of candidates of proposed edge points with respect to a single edge searching direction are detected in a region Ep34 in the figure. In addition, there are some cases where an edge is extracted in a direction which is not perpendicular to the contour of the pattern as shown in a region Ep36 in the figure. A density variation PF2 of the image in the edge searching direction in this case is shown in FIG. 36A. Furthermore, FIG. 36B shows the density variation PF4 of the image when edge searching is carried out in a direction substantially perpendicular to the contour of the pattern. As can be clearly seen from the comparison of both figures with each other, when the edge searching direction is not perpendicular to the contour of the pattern, the waveform of density variation is rendered broader, so that it is easy to include noises. That is, it can be seen that errors are apt to generate in edge detection.
In order to eliminate these problems, for example, the above described Japanese Patent Laid-Open No. 2001-091231 discloses a method for preparing a graphic form exhibiting characteristics of a pattern and for searching edges in a direction perpendicular to the graphic form to acquire contour data. However, it takes a lot of processing time to calculate searching directions for all of sequences of edge points. In addition, when the prepared graphic form exhibiting characteristics of the pattern includes many edges, there is a problem in that it is not possible to skillfully search edges due to the influence of noises which generate from the edges of the prepared graphic form.
Also in the case of the above described threshold method for extracting contour information on the basis of a threshold value, there are following problems in accordance with a pattern to be inspected. For example, if the maximum value (peak) of density in an ROI is set to be 100% and the minimum value (bottom) is set to be 0%, and if a position having a predetermined threshold value (e.g., 50%) is detected as an edge, the peak and bottom positions of a signal waveform in an edge searching direction are clear as shown by, e.g., a density variation (signal waveform) PF6 in FIG. 37A, in a pattern wherein an edge portion perpendicular to a substrate is formed. Therefore, if the ROI is set so as to include the peak and bottom, it is possible to stably detect an edge by the threshold method. However, in the case of a pattern having a taper or the like, there are some cases where the bottom position is not clear as shown by PF8 in FIG. 37B. Therefore, as shown in the comparison of ROI 84 with ROI 86 in FIG. 37B, if the position of an ROI boundary varies, the position of 0% is shifted, so that the position of the edge is shifted to cause a problem in that the result of measurement includes errors.